Breaking up is hard to do (especially on Christmas)

November 2, 2010 in Data,Internet

David McCandless's TED talk on data visualization is excellent -- you can catch it here -- and Mathias Mikkelsen has highlighted a single analysis that investigates when people are most likely to break up (according to Facebook) (Update: original here):

What makes the chart so appealing is how easy it is to understand, despite the lack of scale and minimal labels. I almost wish that the blue comments were missing, as part of the fun with a chart like this is interpreting it without influence.

So, when do people tend to break up? Clearly, the peaks are over spring break and just before Christmas. In fact (as ?if "spring break" wasn't enough of a tip-off), this graph seems heavily skewed toward a still-in-school demographic, as it mirrors features of the academic calendar. Then again, that population is more likely to a) be breaking up and b) putting it in their Facebook status.

The Christmas "blip" is interesting. Merely "being with family" is probably not enough to account for it, or Thanksgiving would probably show a analogous dip. The "Mondays" weekly pattern in the spring is also somewhat amusing.

April Fool's Day illustrates the importance of scrubbing data - without an intelligent review of the data, a researcher might include those points as legitimate though clearly they are not! (or are they?).

Finally, the heightened frequency around Valentine's Day is surprising to me. Perhaps times during which couples spend more time with each other also illuminate the factors that drive them apart (viz: Valentine's Day, the time leading up to Christmas/Thanksgiving, weekends [culminating in Monday breakups]). But so too do periods when couples actually are apart, like vacations. I should note that Valentine's Day is labeled on the graph, which (because of the behavior of the other labels) instinctively leads me to assume it is a "break up day" - even though it is no higher than the days around it. Unlike the other days, it is labeled due to topical relevancy, not as an interesting datapoint. Yet another reason why labels can be misleading.

I would really like to see the opposite chart posted of when couples tend to get together, as well as a comparison of data culled from Facebook statuses and data based on Facebook relationships.

Update: The original chart may be seen here.

Leave a Comment

Previous post:

Next post: